Elon Musk’s TeraFab Enters the EUV Lithography Race with FEL Technology

In the spring of 2026, Musk officially unveiled the TeraFab chip manufacturing initiative. His reasoning was straightforward: SpaceX and Tesla will require at least one terawatt of computing power in the future—a scale exceeding ten times the current global chip supply capacity.

Later that month, he appeared to make an even bigger bet: entering the lithography machine market.

A blogger posted that, according to TeraFab's announcement, Elon Musk appears to be pursuing the FEL (Free Electron Laser) route to disrupt the traditional EUV monopoly.

Elon Musk later posted on social media, “FEL FTW” (FEL will win), which was widely seen as indirect confirmation of this speculation. Combined with TeraFab’s elongated factory design and Musk’s statement, the FEL technology pathway has become the most popular current hypothesis. According to prior plans, TeraFab will establish an integrated chip manufacturing facility capable of producing both logic and memory chips, with all processes—including lithography, packaging, and testing—consolidated within a single factory.

Similarly, in July last year, semiconductor startup xLight announced the completion of a $400 million oversubscribed Series B funding round. The proceeds will be primarily invested in FEL research and development, aiming to break through the physical limits of existing EUV lithography technology and provide critical light source support for mass production of 2nm and more advanced process chips. Notably, in March last year, Intel’s former CEO Pat Gelsinger posted on LinkedIn that he had joined xLight as Executive Chairman.

For a time, the inherent characteristics of FEL technology sparked lively discussions within the industry.

In the entire AI chip supply chain, ASML of the Netherlands is currently the only company in the world capable of manufacturing EUV equipment, holding over 90% of the lithography equipment market share.

The company's EUV lithography machine employs a laser-produced plasma (LPP) EUV source, which operates by directing a 30 kW carbon dioxide laser at tin metal droplets ejected from a nozzle at a rate of 50,000 droplets per second. Each droplet is struck twice (requiring 100,000 laser pulses per second), vaporizing them into plasma and generating 13.5 nm EUV light through transitions between high-energy states of tin ions.

Although LPP technology enabled the commercialization of EUV lithography, its inherent physical limitations are accelerating as process nodes advance.

First, there is the bottleneck of energy conversion efficiency. In the explanation above, a key term is the 13.5nm wavelength. This means that, compared to the 193nm light source used in current mainstream DUV lithography machines, EUV’s light source is only one-fifteenth the wavelength, enabling the etching of much smaller channels on silicon wafers. Currently, ASML primarily uses a carbon dioxide laser from U.S.-based Cymer to excite tin plasma and generate 13.5nm extreme ultraviolet light. Cymer’s technology achieves a conversion efficiency of 5.5% from the laser to the tin plasma. Adding the CO2 laser’s own electrical-to-optical efficiency of approximately 10%, along with transmission losses from the collection mirrors, the actual EUV light utilization from the power grid to the wafer is generally less than 0.5%.

Second is tin debris contamination. During plasma generation, high-speed sputtered tin ions and neutral debris continuously deposit on the surface of the extremely expensive multilayer mirror, leading to reduced reflectivity and shortened lifespan.

Third is the power ceiling. Current LPP-EUV sources have reached a maximum EUV power of approximately 600 W. However, to meet manufacturing requirements for the 2-nanometer node and below, EUV power exceeding 1.5 kilowatts is needed. At present, 500–600 W EUV systems primarily rely on multiple patterning to accumulate photon dosage and compensate for the shortfall in source power.

In February this year, ASML announced plans to increase the production efficiency of its next-generation high-NA EUV lithography machines by 50% before 2030, by introducing a brand-new light source system with a power output of up to 1,000 watts. By 2030, the wafer processing capacity of a single EUV machine will rise from 220 wafers per hour to 330 wafers per hour.

Unlike LPP technology, FEL does not rely on plasma conversion. FEL stands for free-electron laser. The entire light source system begins with an electron gun emitting an initial electron beam, which is then accelerated to near the speed of light by a linear accelerator—advanced designs typically use superconducting linear accelerators. The high-density relativistic electron beam enters an undulator composed of a periodic alternating magnetic field, causing the electrons to oscillate transversely and generate spontaneous radiation. The radiation field continuously modulates the electron beam, prompting the electrons to form micro-bunches with a periodicity matching the radiation wavelength. These micro-bunched electrons emit coherent radiation, creating a positive feedback loop that results in exponential amplification of the radiation intensity. With additional techniques such as seed injection, the system ultimately produces a stable EUV beam.

Therefore, the extreme ultraviolet wavelengths generated by FEL are considered a candidate for next-generation lithography. This wavelength range is shorter than today’s 13.5-nanometer EUV wavelength and falls within the soft X-ray region. According to publicly available information, xLight’s technological goal is to precisely tune the Blue-X band (also known as the “beyond EUV” band) in the 2–7 nanometer range.

Moreover, the entire optical path does not involve tin metal droplet bombardment or plasma sputtering, so no metal debris accumulates within the vacuum chamber of the optical path. The EUV-FEL source can also generate EUV power exceeding 10 kW, enabling it to supply more than 1000 W of EUV power to each of 10 EUV lithography machines simultaneously without causing tin contamination on the Mo/Si mirror surfaces.

Peeling back the surface of the lithography industry reveals three distinct technological trajectories: the incremental evolution of EUV, innovations in EUV light sources, and non-EUV alternatives. All three coexist, but EUV evolution remains the absolute main focus, while the latter two serve more as strategic "moves" in the game.

First camp: ASML’s incremental iterations remain the absolute mainstay.

ASML continues to firmly dominate the mainstream market. In the first quarter, it reported net sales of €8.8 billion and net profit of €2.8 billion; in the second quarter, total net sales reached €9.326 billion with a net profit of €2.918 billion. Meanwhile, ASML has raised its full-year guidance for the second time this year, significantly increasing its 2026 full-year sales forecast to €43–45 billion.

As the leading manufacturer of lithography machines—the most critical equipment in the upstream semiconductor manufacturing process—ASML’s surge in performance reflects the ongoing arms race across the entire technology industry. Giants such as Amazon, Google, and Microsoft have invested hundreds of billions of dollars in infrastructure, triggering massive downstream demand for advanced AI chips. This has accelerated capacity expansion at wafer fabs, including those producing logic and memory chips, driving demand for lithography machines to unprecedented levels.

Capacity expansion is equally aggressive. The company plans to increase capacity by 30% in 2027 based on a 2026 baseline of approximately 65 low-NA EUV tools, and is studying a further 30% capacity increase in 2028. Meanwhile, based on a 2026 baseline of approximately 130 immersion DUV tools, the company plans to increase capacity by 30% in 2027 and is studying an additional 30% capacity increase in 2028.

High-NA EUV lithography machines, ASML’s next-generation flagship, feature a 0.55 numerical aperture optical system capable of achieving 8-nanometer resolution, enabling production processes at 3 nanometers and below, while laying the technological foundation for the 1-nanometer node. This equipment increases circuit patterning precision by 1.7 times through single-exposure imaging, improves image contrast by 40%, and achieves transistor density 2.9 times higher than previous systems, effectively reducing chip power consumption and enhancing processing speed.

However, due to the high cost of a single High-NA EUV machine—approximately $400 million, nearly double that of traditional EUV lithography systems—and the significant technical challenges involved in adapting and integrating it into production lines, the adoption of High-NA EUV has not been ideal. Zhang Xiaoqiang, Senior Vice President and Co-Head of Operations responsible for business development and global operations at TSMC, revealed at a press conference ahead of the annual technology forum that the company currently has no plans to deploy ASML’s High-NA EUV equipment, which is specifically designed for next-generation processors.

Second camp: Light source innovation—precisely targeting ASML’s “heart.”

This is the FEL route chosen by Musk’s TeraFab and xLight: rather than directly challenging ASML’s dominance in optical lithography systems, they are targeting breakthroughs in the light source component. This means chip manufacturers can achieve significant improvements in capacity and cost without replacing their existing lithography, etching, deposition, or inspection equipment—simply by swapping out the light source system. This “plug-and-play” upgrade path is highly attractive to wafer fabs.

xLight claims that its FEL system delivers more than four times the power of existing systems; deploying xLight FEL in current U.S. fabs can increase production efficiency by 50% and eliminate the need for consumables such as tin or hydrogen, while deploying it in new fabs can boost production efficiency by 100%. This will enable manufacturers to produce chips with smaller feature sizes and higher efficiency, advancing next-generation lithography technology.

If the light source could be replaced independently, ASML’s bargaining power would be structurally weakened. However, it is worth noting that ASML considered the FEL EUV light source pathway over a decade ago but ultimately deemed it too risky, opting instead for the LPP EUV light source. Therefore, whether the mass production challenges of FEL can be overcome—and when—may still require further time and investigation.

In addition, there are other light source pathways. For example, Substrate, a San Francisco-based startup founded in 2022, has chosen a particle accelerator-based X-ray lithography approach. China is also advancing its domestic EUV light source technology. According to public information, multiple domestic teams are exploring various technical pathways, including reverse engineering of existing LPP technology, with the goal of achieving breakthroughs between 2028 and 2030.

Among them, institutions such as Harbin Institute of Technology have attempted to develop a lithography scheme based on laser-induced discharge plasma (LDP), which works by evaporating tin between electrodes and exciting plasma through high-voltage discharge. This structure is simpler and more compact than LPP, but its light emission power density is limited, raising questions about its viability for mass production.

The third camp: Non-traditional approaches that completely bypass EUV are growing in the shadows.

Nanoimprint lithography (NIL) is the most commercially advanced among these technologies. By directly stamping patterns using a physical mold, NIL eliminates the need for complex optical systems and light sources, resulting in significantly lower equipment costs and power consumption compared to EUV. Japanese manufacturer Canon has already driven the mass production application of NIL in memory chips. Although its resolution cannot yet match that of High-NA EUV, NIL has demonstrated cost competitiveness in the memory chip market, where linewidth requirements are relatively lenient.

Electron beam lithography (EBL) takes a fundamentally different approach. Essentially a direct-write technique, EBL uses a focused electron beam to expose resist point by point, precisely defining patterns through electromagnetic control. This method does not require a mask, making it highly advantageous during research and development phases involving frequent design iterations—particularly suited for applications such as quantum devices, novel material structures, prototype chips, and mask fabrication.

However, so far, electron beam lithography has remained primarily in the realm of research and small-scale applications, not due to resolution, but because of efficiency. Electron beam direct writing is a serial exposure process; even though individual point accuracy is very high, overall throughput remains limited, which is unacceptable in wafer-level mass production. Yet, during the R&D phase, this “slow” characteristic offers exceptional flexibility. For research teams that frequently need to modify layouts, validate physical models, or explore new device structures, eliminating the mask fabrication process is often more valuable than increasing exposure speed.

Today, EUV is more like a seasoned veteran that has been running for a long time. While technical bottlenecks are real, so too is the ecosystem ASML has built around it over two decades. For a new entrant to join the race, simply running faster isn’t enough—they must convince the entire track to change its rules.

Musk’s TeraFab project is essentially a high-stakes gamble: betting that FEL can move from the lab to the wafer fab, that “plug-and-play” light source replacements can bypass ASML’s patent barriers, and that a terawatt of computing demand will be sufficient to support an entirely new supply chain.

The path forward for the next-generation lithography technology may not be far off. But one thing is certain—when Musk posts “FEL FTW” on social media, the lithography machine business is no longer just ASML’s game.

This article is from the WeChat public account "Semiconductor Industry纵横" (ID: ICViews), author: Feng Ning.